femorin/wrata

Wrata is a Python script for Windows and Linux that takes a screenshot when you press a hotkey, sends the image to an AI model, and delivers the AI's answer to your Telegram account. The intended use described in the README is answering test questions by capturing what is on the screen. The README is written in Russian. When you start the program, a small window asks which AI model to use. The Mistral option sends the screenshot to Pixtral to extract the question from the image, then passes it to Mistral Large for an answer, which arrives in Telegram as a short reply. The Gemini option uses OpenRouter to send the screenshot to Gemini for a detailed response, then uses Mistral to compress the answer down to one line before sending it to Telegram. Setup requires a Telegram bot token and your Telegram chat ID, plus API keys for whichever model you plan to use. Mistral offers a free tier without requiring a payment method. OpenRouter is needed only if you want the Gemini mode. You fill these values into a configuration file before running the program. On Windows the program runs silently using a .vbs launcher file with no visible console window or system tray icon. On Linux it requires the X11 display server, Wayland is not supported natively, though the README notes a workaround for running it under Wayland. The hotkey that triggers the screenshot and the hotkey that exits the program are both configurable in the same configuration file. There is a separate repository for macOS users, linked in the README.